A total score for the CERAD

neuropsychological battery
M.J. Chandler, PhD; L.H. Lacritz, PhD; L.S. Hynan, PhD; H.D. Barnard, BA; G. Allen, PhD;
M. Deschner, PhD; M.F. Weiner, MD; and C.M. Cullum, PhD

Abstract—Objective: To develop a total or composite score for the Consortium to Establish a Registry for Alzheimer’s
Disease (CERAD) neuropsychological battery. Method: CERAD total scores were obtained by summing scores from the
individual CERAD subtests (excluding the Mini-Mental State Examination [MMSE]) into a total composite (maximum
score ⫽ 100). The method of tabulating the total score was constructed using normal controls (NCs; n ⫽ 424) and patients
with AD (n ⫽ 835) from the CERAD registry database. The utility of the total score was further tested in independent
samples of mild AD (n ⫽ 95), mild cognitive impairment (MCI; n ⫽ 60), and NC (n ⫽ 95) subjects. Results: The CERAD
total score was highly accurate in differentiating NC and AD subjects in the CERAD registry. Age, gender, and education
effects were observed, and demographic correction scores were derived through multiple regression analysis. Demograph-
ically corrected CERAD total scores showed excellent test–retest reliability across samples (r ⫽ 0.95) and were highly
correlated with the MMSE (r ⫽ 0.89) and Clinical Dementia Rating Scale (r ⫽ ⫺0.83) in mixed AD and NC samples and
with the Blessed Dementia Rating Scale in an AD sample (r ⫽ ⫺0.40). The CERAD total score was highly accurate in
differentiating independent samples of NC, MCI, and AD subjects. Conclusion: Results provide support for the validity of
a Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) total score that can be used along with the
normative data to provide an index of overall level of cognitive functioning from the CERAD neuropsychological battery.
NEUROLOGY 2005;65:102–106

The Consortium to Establish a Registry for Alzhei-
mer’s Disease (CERAD) neuropsychological battery
is composed of five subtests derived from previously
established cognitive tests (Animal Naming,1 Modi-
fied Boston Naming Test [BNT],2 Mini-Mental State
Examination [MMSE],3 Constructional Praxis,4 and
Word List Memory).5 These subtests have been found
to be valid and reliable measures of cognition in nor-
mal aging and in Alzheimer disease (AD).6 Although
grant funding for the original CERAD project ended
in 1995,7 several AD research centers continue to use
the CERAD neuropsychological battery, as do many
major medical centers, universities, and private clin-
ics in the United States and abroad, both clinically
and in epidemiologic studies.8 Currently, there is no
summary score for the CERAD neuropsychological
battery, limiting reporting of cognitive ability and
level of dementia to a description of multiple data
points across cognitive domains with no simple
means of depicting overall level of cognitive impair-
ment. We sought to develop a method for tabulating
a total score for the CERAD neuropsychological bat-
Additional material related to this article can be found on the Neurology
Web site. Go to www.neurology.org and scroll down the Table of Con-
tents for the July 12 issue to find the title link for this article.
tery that would provide an easy-to-use, normative-
based summary score of global cognitive performance
that could be used to identify level of cognitive im-
pairment and aid in differentiating AD, mild cogni-
tive impairment (MCI), and normal aging.
Methods. Four hundred twenty-four normal control (NC) and
835 AD subjects from the CERAD registry were included. Controls
consisted of spouses of CERAD AD subjects and community volun-
teers. AD subjects were diagnosed with probable or possible AD
using National Institute of Neurological and Communication Dis-
orders and Stroke/Alzheimer’s Disease and Related Disorders As-
sociation (NINCDS-ADRDA)9 guidelines. As a limited number of
minority control subjects were collected for the registry, only Cau-
casian subjects were included, in keeping with previously reported
normative data from the CERAD registry.10 All subjects were age
50 or older, English speaking, not institutionalized, and free of
co-morbid conditions that could affect cognition.11 All CERAD reg-
istry controls received a Clinical Dementia Rating Scale (CDR)
score of 0.12 AD subjects had a caregiver or informant who could
provide history on the subject. Subjects from the Clinic for Alzhei-
mer’s and Related Diseases (ADC) at the University of Texas
Southwestern (UTSW) Medical Center at Dallas were excluded
from the CERAD data set to prevent overlap with the UTSW ADC
sample.
Three subject groups were selected from the UTSW ADC: 1) 95
healthy Caucasian subjects recruited from the surrounding com-
munity who met the same inclusion criteria as CERAD registry
NC subjects; 2) a random sample of 95 subjects with probable AD
per NINCDS-ADRDA criteria of mild dementia severity, as de-
fined by a CDR score of 1, selected from 304 Caucasian subjects;
and 3) 60 Caucasian subjects diagnosed with amnestic MCI using

From the Department of Psychiatry and Psychology (Dr. Chandler), Mayo Clinic, Rochester, MN, and Department of Psychiatry (Drs. Chandler, Lacritz,
Hynan, Barnard, Allen, Weiner, and Cullum), Center for Biostatistics and Clinical Science (Dr. Hynan), Department of Anesthesiology and Pain Manage-
ment (Dr. Deschner), and Department of Neurology (Drs. Weiner and Cullum), University of Texas Southwestern Medical Center at Dallas. Supported in
part by NIH, NIA grant P30AG12300. Presented in part at the International Neuropsychological Society Meetings in Baltimore, MD (2004), and St. Louis,
MO (2005).
Received August 20, 2004. Accepted in final form March 18, 2005.
Address correspondence and reprint requests to Dr. C.M. Cullum, University of Texas Southwestern Medical Center at Dallas, 5323 Harry Hines Blvd.,
Dallas, TX 75235-8846; e-mail: munro.cullum@utsouthwestern.edu

102 Copyright © 2005 by AAN Enterprises, Inc.

Table 1 CERAD total score tabulation
Subtest Maximum points possible
Verbal Fluency 24*
Modified BNT 15
Word List Learning 30
Constructional Praxis 11
Word List Recall 10
Word List Recognition
Discriminability 10†
Total score 100
* Verbal Fluency does not have a ceiling when administered us-
ing standard instructions as part of the Consortium to Estab-
lish a Registry for Alzheimer’s Disease (CERAD)
neuropsychological battery. For calculation purposes, a cap of
24 was placed on Verbal Fluency, which represents 1 SD above
the normal aging population mean.10
† Recognition Discriminability was calculated by subtracting the
number of false positives from the number of true positives.
BNT ⫽ Boston Naming Test.
Mayo Clinic criteria.13 Although information about CERAD neuro-
psychological battery performance was available at the time of
diagnosis, diagnosis was made by an interdisciplinary team and
not dependent upon performance on the CERAD neuropsychologi-
cal battery.
All CERAD registry subjects were administered the CERAD
clinical assessment and neuropsychological battery, including the
CDR. Additionally, the Blessed Dementia Rating Scale 14 (BDRS)
was administered to all CERAD registry AD subjects. UTSW ADC
subjects were administered the CERAD neuropsychological bat-
tery, and all AD subjects received CDR ratings. Standard admin-
istration of the CERAD battery has been described previously.11
All subjects (or legal guardians) gave informed consent prior to
testing.
Four methods of constructing a total score from the CERAD
registry NC sample were explored: 1) a subtest addition method;
2) a method in which additional memory variables (Word List
True Positive Recognition, True Negative Recognition, Savings
Score, and Intrusion Score) were added to the subtest totals; 3) a
z-score transformation method that provided equal weight to each
subtest; and 4) a z-score transformation method that provided
equal weight to each cognitive domain. In each case, the MMSE
was excluded because of its global nature and so that comparisons
with this well-established, cognitive screening measure could be
made.
Receiver operating characteristic (ROC) curve analysis re-
vealed that all four methods of tabulating the CERAD total score
were highly accurate in differentiating NC from AD subjects (area
under the curve [AUC] ranged from 0.996 to 0.997; p ⬍ 0.0001).
Therefore, as differential weighting of the subtests or cognitive
domains did not change the accuracy of differential diagnosis, the
easiest tabulation procedure (Method 1; computed by simply sum-
ming subtest scores) was selected as the preferred CERAD total
score tabulation method. The subtest summation procedure for
tabulating the CERAD total score is presented in table 1.
The addition of these measures resulted in a total score with a
maximum of 100 points composed of 39% language (Verbal Flu-
ency and Modified BNT), 30% learning (Word List Learning), 11%
construction (Constructional Praxis), and 20% memory (Word List
Recall and Recognition Discriminability). Praxis recall15 was not
included in the summation of this score as it was not included in
the original CERAD registry data, and its reliability and validity
have not yet been adequately established.
Age, education, and gender effects on the CERAD neuropsy-
chological battery and its subtests have been highlighted in the
literature.10,16 Therefore, a multiple regression formula was used
in the total score tabulation in the normative sample (CERAD
Table 2 Characteristics of CERAD registry samples
Characteristic NC, mean (SD) AD, mean (SD)
Age* 68.3 (8.0) 71.9 (7.9)
Education* 13.9 (3.0) 12.6 (3.5)
MMSE* 28.9 (1.4) 17.5 (5.6)
% Male* 34.9 44.3
% CDR ⫽ 0–0.5 100.0 3.5
% CDR ⫽ 1 0.0 53.1
% CDR ⫽ 2 0.0 38.6
% CDR ⬎ 3 0.0 4.9
BDRS N/A 4.5 (2.5)
* p ⬍ 0.001.
CERAD ⫽ Consortium to Establish a Registry for Alzheimer’s
Disease; NC ⫽ normal control; AD ⫽ Alzheimer disease;
MMSE ⫽ Mini-Mental State Examination; CDR ⫽ Clinical De-
mentia Rating Scale; BDRS ⫽ Blessed Dementia Rating Scale.
registry NC sample), with weights assigned for each significant
demographic variable.
One-month test–retest correlations were determined using a
subset of 240 AD subjects and 300 NC subjects from the CERAD
registry for whom the CERAD neuropsychological battery was
readministered. Convergent validity of the CERAD total score was
determined by examining correlations of the total score with the
MMSE and CDR in a mixed sample of the CERAD registry control
and AD samples to allow for a maximum range of scores. Correla-
tions between the CERAD total score and BDRS were examined in
the AD sample alone, as the BDRS was not administered to NCs.
ROC analyses were performed using the CERAD total score,
MMSE, and Word List Recall to determine their accuracy in dis-
tinguishing among AD, MCI, and NC samples. Accuracy scores
were then compared using a z-score transformation17 to determine
which test obtained the higher score. Likelihood ratios were also
calculated for cut-off scores.18
Results. Total score tabulation in CERAD registry. Char-
acteristics of the NC and AD groups from the CERAD regis-
try samples are in table 2. NC subjects were younger
(t[1,257] ⫽ 7.5, p ⬍ 0.001) and more educated (t[959.1] ⫽
⫺6.6, p ⬍ 0.001) than AD subjects. Actual mean differences
for age and education between the groups were small, but
small to moderate effect sizes were observed (age d ⫽ 0.44
and education d ⫽ 0.37). There was also a larger proportion
of women in the NC group than the AD sample (␹2[1, n ⫽
1,259] ⫽ 10.3).
Summing the raw scores on the subtests of the CERAD
neuropsychological battery resulted in a mean total score
of 79.7 (SD ⫽ 9.0, range ⫽ 52 to 98) in the NC group and
37.3 (SD ⫽ 13.7, range ⫽ 2 to 77) in AD subjects. Age
(r[422] ⫽ ⫺0.35) and education (r[422] ⫽ 0.33) were corre-
lated with the total scores in the NC group (p ⬍ 0.001) and
in a mixed sample of NC and AD registry subjects, age
(r [1,257] ⫽ ⫺0.25) and education (r[1,257] ⫽ 0.26, p ⬍
0.001). Specifically, older subjects performed worse than
younger subjects, and more educated subjects outper-
formed less educated subjects. Women (mean ⫽ 80.9, SD ⫽
8.7) outperformed men (mean ⫽ 77.4, SD ⫽ 9.0) in the regis-
try NC sample, even after controlling for age and education
(F[1,420] ⫽ 12.3, p ⫽ 0.001). In contrast, men (mean ⫽
38.6, SD ⫽ 13.5) outperformed women (mean ⫽ 36.1, SD ⫽
13.7) in the AD sample, but this appears to be due to
women being less educated, as gender was no longer a
July (1 of 2) 2005 NEUROLOGY 65 103
factor after controlling for education (F[1,832] ⫽ 3.0, p ⫽ Table 3 Demographic and cognitive characteristics of the NC,
0.09). MCI, and AD samples
With use of multiple regression analysis, the linear NC, mean MCI, mean AD, mean
combination of age, education, and gender was related to Characteristic (SD) (SD) (SD)
the total score (F[3,420] ⫽ 42.3, p ⬍ 0.0001) in the NC
sample. The sample multiple correlation coefficient was Age, y 74.2 (7.2) 72.8 (7.5) 74.4 (7.4)
R ⫽ 0.48, indicating that 23.2% of the variance of the total Education, y 15.3 (2.4)*† 14.8 (2.8)† 13.6 (2.8)
score could be accounted for by the linear combination of % male 40.0 48.3 38.9
these demographic variables (unique variance for age ⫽
CERAD total
8.0%, education ⫽ 9.1%, and gender ⫽ 2.2%). The demo-
score
graphic correction regression formula was as follows: raw
score - (⫺0.324 * age ⫹ 0.897 * education ⫺ 2.858 * gen- Uncorrected 79.2 (9.6)*† 68.0 (10.0)† 49.5 (11.7)
der). Mean corrected total scores were 90.4 (SD ⫽ 7.8) in Corrected 88.4 (9.3)*† 76.9 (8.9)† 60.2 (11.9)
the NC group and 50.5 (SD ⫽ 13.7) in AD subjects. Demo- MMSE 28.6 (1.3)*† 27.5 (1.8)† 21.1 (4.1)
graphically corrected and uncorrected total scores were
Word List 6.9 (1.9)*† 4.0 (2.5)† 1.5 (1.6)
higher in the NC sample (t[1,240.5] ⫽ ⫺65.5, p ⬍ 0.0001)
Recall
than the AD sample (t[1,180.1] ⫽ ⫺66.0, p ⬍ 0.0001). The
regression formula was used to calculate correction factor BDRS — 2.3 (1.4)† 3.5 (1.7)
scores by age, education, and gender that can be added to * Scores differ from the MCI group (p ⬍ 0.001).
the raw total score for clinical use (see table E-1 on the † Scores differ from the AD group (p ⬍ 0.001).
Neurology Web site at www.neurology.org). Normative
data in the form of T scores are provided for the demo- NC ⫽ normal control; MCI ⫽ mild cognitive impairment; AD ⫽
Alzheimer disease; CERAD ⫽ Consortium to Establish a Regis-
graphically corrected CERAD total scores in table E-2.
try for Alzheimer’s Disease; MMSE ⫽ Mini-Mental State Exami-
Test–retest reliability. One-month test–retest reliabili-
nation; BDRS ⫽ Blessed Dementia Rating Scale.
ties for the uncorrected (NC r[298] ⫽ 0.78; AD r[238] ⫽
0.89) and corrected (NC r[298] ⫽ 0.70; AD r[238] ⫽ 0.90)
CERAD total scores were very similar (p ⬍ 0.0001), dem- 179.1, p ⬍ 0.0001), with NC subjects (mean ⫽ 88.4, SD ⫽
onstrating the high reliability of the total score. 9.3) outperforming MCI subjects (mean ⫽ 76.9, SD ⫽ 8.9)
Convergent validity. As with the CERAD total score, and MCI outperforming AD subjects (mean ⫽ 60.2, SD ⫽
MMSE scores were higher in the NC group (mean ⫽ 28.9, 11.9). CERAD total scores were correlated with MMSE,
SD ⫽ 1.4) than in the AD sample (mean ⫽ 17.5, SD ⫽ 5.6) BDRS, and CDR stage scores as well as the subtests of the
(t[1,014.6] ⫽ ⫺54.9, p ⬍ 0.0001). The MMSE was corre- CERAD neuropsychological battery in the UTSW ADC
lated with the total score in the mixed AD and NC samples samples (all p ⬍ 0.0001). Among the subtests, Word List
(r[1,258] ⫽ 0.89, p ⬍ 0.0001). The CDR (NC mean ⫽ 0.0, Learning total was most highly correlated with the
SD ⫽ 0.0; AD mean ⫽ 1.5, SD ⫽ 0.6) was correlated with CERAD total score (r[248] ⫽ 0.90, p ⬍ 0.0001), and Con-
the total score in the mixed sample (r[1,259] ⫽ ⫺0.83, p ⬍ structional Praxis had the weakest, albeit still robust, cor-
0.0001). The BDRS score (AD mean ⫽ 4.5, SD ⫽ 2.5) was relation (r[248] ⫽ 0.50, p ⬍ 0.0001). Correlations among
correlated with the total score in the AD sample (r[831] ⫽
⫺0.40, p ⬍ 0.0001). Table 4 Correlations of CERAD total score with MMSE, BDRS,
Tabulation of total scores in UTSW ADC samples. and CERAD subtests across all subjects
Characteristics of the NC, MCI, and AD groups from the
UTSW ADC samples are presented in table 3. No signifi- Parameter CERAD total score, r
cant differences in age (F[2,247] ⫽ 0.90, p ⫽ 0.41) or gen- MMSE† 0.76*
der (␹2[2, n ⫽ 250] ⫽ 1.5, p ⫽ 0.47) were present across
BDRS‡ ⫺0.32*
groups. The three groups were different for education
(F[247] ⫽ 10.0, p ⬍ 0.0001), with AD subjects having less Verbal Fluency 0.81*
education than the MCI (p ⫽ 0.02, d ⫽ 0.41) and NC (p ⬍ Modified BNT 0.61*
0.001, d ⫽ 0.62) groups, with moderate effect sizes. Constructional Praxis 0.50*
Uncorrected CERAD total scores in the UTSW ADC NC
Word List Learned 0.90*
sample (mean ⫽ 79.2, SD ⫽ 9.6) were highly similar to
those in the CERAD registry sample (mean ⫽ 79.7, SD ⫽ Word List Recall 0.86*
9.0) and highly correlated with demographically corrected Word List Discriminability 0.72*
total scores (r[95] ⫽ 0.93, p ⬍ 0.0001). The demographic
correction factors (derived by applying the CERAD registry n ⫽ 250. NC subjects were not routinely given the BDRS, result-
ing in lower n values.
regression weights in the UTSW ADC NC sample) were
correlated with the raw total scores in the UTSW ADC NC * p ⬍ 0.001.
sample (r[95] ⫽ 0.27, p ⫽ 0.008), providing support that † n ⫽ 249 (1 MCI subject did not have the MMSE).
these two samples are similar. Thus, demographic correc- ‡ n ⫽ 130 (AD ⫽ 87, MCI ⫽ 43, NC ⫽ 0).
tions in the UTSW ADC NC sample were based on the
CERAD ⫽ Consortium to Establish a Registry for Alzheimer’s
CERAD registry regression formula weights, resulting in a Disease; MMSE ⫽ Mini-Mental State Examination; BDRS ⫽
mean total score of 88.4 (SD ⫽ 9.3). Blessed Dementia Rating Scale; BNT ⫽ Boston Naming Test;
Demographically corrected CERAD total scores were NC ⫽ normal control; AD, Alzheimer disease; MCI ⫽ mild cogni-
different among the AD, MCI, and NC groups (F[2,247] ⫽ tive impairment.
104 NEUROLOGY 65 July (1 of 2) 2005
Table 5 AUCs and cut-off scores of CERAD total score, MMSE, score that were explored, no method was signifi-
and Word List Recall in AD, MCI, and NC samples cantly better than the others in discriminating NC
Parameter NC and AD NC and MCI MCI and AD and AD subjects. This is consistent with the observa-
tion that assigning differing weights to variables
CERAD total score makes little to no difference in outcome when vari-
AUC 0.958 0.823* 0.882†‡ ables are positively intercorrelated,19 as are the
Cut-off 77.0 85.1 68.5 subtests of the CERAD neuropsychological battery.11
Sensitivity/specificity 93.7/92.6 81.4/72.6 80.0/81.4 Thus, the subtest addition method was selected as
the preferred total score method because it was the
Likelihood ratio 12.7 3.0 4.3
most straightforward and easiest to tabulate.
MMSE total score One-month test–retest reliability for the corrected
AUC 0.980 0.692 0.938 CERAD total score was excellent across samples (r ⫽
Cut-off 26.5 28.5 25.5 0.95 in a mixed AD and NC sample), suggesting that
Sensitivity/specificity 92.6/94.7 66.1/61.1 88.4/83.1 the CERAD total score is a reliable measure of cogni-
tive functioning in AD and NC groups. The NC sample
Likelihood ratio 17.5 1.7 5.2
demonstrated lower but still good test–retest correla-
Word List Recall
tions (r ⫽ 0.70 for corrected total scores), likely due to a
AUC 0.969 0.823 0.816 more restricted range of total scores in this group com-
Cut-off 4.5 6.5 2.5 pared with the AD group. Whereas ceiling and floor
Sensitivity/specificity 92.6/89.5 81.4/63.2 81.1/67.8 effects have been noted for some of the individual
Likelihood ratio 8.8 2.2 2.5
subtests of the CERAD neuropsychological battery,11
ceiling and floor effects were not found for the total
CERAD total scores are demographically corrected. score in the AD or NC samples, reflecting an improve-
ment over individual subtests alone.
* Score was better than MMSE AUC (p ⫽ 0.004).
† Score was worse than MMSE AUC (p ⫽ 0.01). The utility of the total score was further sup-
‡ Score was better than Word List Recall AUC (p ⬍ 0.007). ported as this measure demonstrated the most con-
sistent differential diagnostic ability across MCI,
AUC ⫽ area under the curve; CERAD ⫽ Consortium to Establish
AD, and NC subgroups compared with the MMSE,
a Registry for Alzheimer’s Disease; MMSE ⫽ Mini-Mental State
Examination; AD ⫽ Alzheimer disease; MCI ⫽ mild cognitive which had difficulty differentiating MCI from NC
impairment; NC ⫽ normal control. samples, and the Word List Recall subtest, which
was not as effective in differentiating MCI from AD.
On the MMSE, failure to distinguish MCI from NC
these measures with the CERAD total score are provided subjects appeared to be due to ceiling effects, as 92%
in table 4. of subjects scored ⱖ25 in the MCI sample and 95%
Diagnostic utility in AD, MCI, and normal aging. ROC scored ⱖ27 in the NC sample. From a clinical stand-
curve analysis revealed that the CERAD total score was point, use of the CERAD total score would result in
highly accurate in differentiating NC and AD subjects in fewer false negatives than the MMSE when distin-
the UTSW ADC sample (p ⬍ 0.0001). AUCs, cut-off scores, guishing MCI and normal aging, allowing more indi-
and corresponding sensitivity/specificity and likelihood ra- viduals with MCI to be correctly identified and
tios are presented in table 5. provided with further testing and treatment as indi-
The CERAD total score was similar to the MMSE (z ⫽ cated. As the CERAD neuropsychological battery is
1.61, p ⬎ 0.06) and Word List Recall (z ⫽ 0.92, p ⬎ 0.17) in more comprehensive than the MMSE and provides a
its ability to differentiate NC and AD subjects. The wider range of scores across cognitive domains, this
CERAD total score was better at differentiating NC from
also makes the testing battery somewhat longer and
MCI subjects than the MMSE (z ⫽ 2.65, p ⫽ 0.004). How-
requires more training to administer. Nevertheless,
ever, the MMSE did slightly better than the CERAD total
score in distinguishing MCI and AD subjects (z ⫽ 2.20, p ⫽
the specificity of the CERAD total score to MCI and
0.01). The CERAD total score was superior to Word List AD, the increased information provided by the
Recall in differentiating MCI from AD subjects (z ⫽ 2.46, CERAD’s breadth, and the fact that the CERAD was
p ⬍ 0.007). However, the CERAD total score was similar to specifically designed for AD diagnosis whereas the
the Word List Recall subtest in distinguishing NC from MMSE was not may make the CERAD total score a
MCI subjects (z ⫽ 0, p ⫽ 0.5). better clinical choice for AD assessment than the
MMSE.
Discussion. Until now, reporting of performance Word List Recall is a sensitive test to AD, but
on the CERAD neuropsychological battery has been floor effects have been reported to restrict its utility
limited to a description of multiple data points across in differentiating stages of AD.11 This limitation was
cognitive domains, with no index of overall cognitive apparent in the current study when differentiating
disability or composite score. Summing the subtests MCI from AD, albeit with MCI scores not restricted to
resulted in CERAD total scores that differentiated the degree found in the AD sample (MCI median ⫽ 3,
mild AD, MCI, and NC samples with high accuracy. AD median ⫽ 1, NC median ⫽ 7). These findings
Of the four different methods of tabulating a total support the added utility of the CERAD total score
July (1 of 2) 2005 NEUROLOGY 65 105
over the use of the Word List Recall subtest when
differentiating impaired groups (i.e., MCI and AD).
Cut-off scores with corresponding sensitivity and
specificity ratings for normal aging, MCI, and mild
AD were provided for guidance in diagnosing individ-
uals into these categories more accurately. The cut-
off scores were not 100% accurate in differentiating
the groups, as may be expected in trying to distin-
guish clinical diagnoses based on a single measure.
It should be noted that sensitivity/specificity ratios
are limited by the characteristics of the groups being
compared.20 Therefore, as the current study com-
pared mild AD, MCI, and NC subjects, sensitivity
and specificity would vary if scores for patients with
moderate or severe AD were examined. Neverthe-
less, the high sensitivity and specificity of the
CERAD total score suggest that this score can be an
effective tool when used as a screening measure in
which further clinical evaluation will be performed
for diagnosis. Interested clinicians can use the listed
sensitivity and specificity ratios to calculate positive
and negative predictive values for diagnosis based on
the base rates of AD, MCI, and NC in their patient
population. In addition, likelihood ratios have been
provided to estimate an individual’s risk of having a
diagnosis based on the cut-off scores.
The generalizability of the CERAD total score
could be improved by addressing the small number
of individuals with low education and lack of minor-
ity subjects in the current samples. Also, it should be
noted that the CERAD total score was designed spe-
cifically to differentiate AD, MCI, and NC samples.
Therefore, the utility of the total score in differential
diagnosis of AD from other dementias will require
further study.
A next step in the evolution of the CERAD total
score would be to explore its utility in the progres-
sion of AD dementia. Cut-off scores for mild, moder-
ate, and severe dementia could be established and
normative data produced for each level of severity.
Correlations between the total score and measures of
brain pathology could be examined to establish the
score’s validity in staging disease progression, as has
been done in the MMSE, DRS, and BDRS.21 Further,
the ability of the total score to classify AD subjects
106 NEUROLOGY 65 July (1 of 2) 2005
into each of these levels of impairment could be com-
pared with other measures (e.g., MMSE).
References
1. Isaacs B, Kennie A. The Set Test as an aid to the detection of dementia
in old people. Br J Psychiatry 1973;123:467–470.
2. Kaplan E, Goodglass H, Weintraub S. The Boston Naming Test. Boston:
Veterans Administration Medical Center, 1978.
3. Folstein M, Folstein S, McHugh P. “Mini-Mental State.” A practical
method for grading the cognitive state of patients for the clinician. J
Psychiatr Res 1975;12:189–198.
4. Rosen W, Mohs R, Davis K. A new rating scale for Alzheimer’s disease.
Am J Psychiatry 1984;141:1356–1364.
5. Atkinson R, Shiffrin R. The control of short-term memory. Sci Am
1971;225:82–90.
6. Welsh-Bohmer K, Mohs R. Neuropsychological assessment of Alzhei-
mer’s disease. Neurology 1997;49(suppl 3):S11–S13.
7. Buckholtz N. Background and purpose of the Consortium to Establish a
Registry for Alzheimer’s Disease. Neurology 1997;49(suppl 3):S1.
8. Heyman A, Fillenbaum G. Overview: clinical sites, case material, and
special studies. Neurology 1997;49(suppl 3):S2–S6.
9. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan E.
Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-
ADRDA Work Group under the auspices of Department of Health and
Human Services Task Force on Alzheimer’s Disease. Neurology 1984;
34:939–944.
10. Welsh K, Butters N, Mohs R. et al. The Consortium to Establish a
Registry for Alzheimer’s Disease (CERAD). Part V. A normative study
of the neuropsychological battery. Neurology 1994;44:609–614.
11. Morris J, Heyman A, Mohs R. et al. The Consortium to Establish a
Registry for Alzheimer’s Disease (CERAD). Part I. Clinical and neuro-
psychological assessment of Alzheimer’s disease. Neurology. 1989;39:
1159–1165.
12. Berg L. Clinical Dementia Rating (CDR). Psychopharmacol Bull 1988;
24:637–639.
13. Petersen R, Smith G, Waring S, Ivnik R, Tangalos E, Kokmen E. Mild
cognitive impairment: clinical characterization and outcome. Arch Neu-
rol 1999;56:303–308.
14. Blessed G, Tomlinson B, Roth M. The association between quantitative
measures of dementia and of senile change in the cerebral grey matter
of elderly subjects. Br J Psychiatry 1968;114:797–811.
15. Yuseph R, Vanderploeg R, Dupree L, Cohen D. Preliminary normative
data for the delayed recall of a constructional praxis task as part of the
CERAD battery. Arch Clin Neuropsychol 1997;12:433.
16. Fillenbaum G, Heyman A, Huber M, Ganguli M, Unverzagt F. Perfor-
mance of elderly African American and White community residents on
the CERAD Neuropsychological Battery. J Int Neuropsychol Soc 2001;
7:502–509.
17. Hanley J, McNeil B. A method of comparing the areas under receiver
operating characteristic curves derived from the same cases. Radiology
1983;148:839–843.
18. Radack K, Rouan G, Hedges J. The likelihood ratio. An improved mea-
sure for reporting and evaluating diagnostic test results. Arch Pathol
Lab Med 1986;110:689–693.
19. Wainer H. Estimating coefficients in linear models: it doesn’t make no
nevermind. Psychol Bull 1976;83:213–217.
20. Smith G, Cerhan J, Ivnik R. Diagnostic validity. In: Tulsky D, Sak-
lofske D, Heaton R, Bornstein R, Ledbetter M, eds. Clinical interpreta-
tion of the WAIS-III and WMS-III. New York: Academic Press, 2003:
273–301.
21. Terry R, Masliah E, Salmon D, et al. Physical basis of cognitive alter-
ations in Alzheimer’s disease: synapse loss is the major correlate of
cognitive impairment. Ann Neurol 1991;30:572–580.
 A total score for the CERAD neuropsychological battery
M. J. Chandler, L. H. Lacritz, L. S. Hynan, et al.
Neurology 2005;65;102-106
DOI 10.1212/01.wnl.0000167607.63000.38

This information is current as of July 11, 2005

Updated Information & including high resolution figures, can be found at:
Services http://www.neurology.org/content/65/1/102.full.html

Supplementary Material Supplementary material can be found at:

http://www.neurology.org/content/suppl/2005/07/08/65.1.102.DC1.htm
l
References This article cites 19 articles, 7 of which you can access for free at:
http://www.neurology.org/content/65/1/102.full.html##ref-list-1
Citations This article has been cited by 8 HighWire-hosted articles:
http://www.neurology.org/content/65/1/102.full.html##otherarticles
Subspecialty Collections This article, along with others on similar topics, appears in the
following collection(s):
All Neuropsychology/Behavior
http://www.neurology.org//cgi/collection/all_neuropsychology_behavio
r
Alzheimer's disease
http://www.neurology.org//cgi/collection/alzheimers_disease
Clinical trials Methodology/study design
http://www.neurology.org//cgi/collection/clinical_trials_methodology_
study_design_
Permissions & Licensing Information about reproducing this article in parts (figures,tables) or in
its entirety can be found online at:
http://www.neurology.org/misc/about.xhtml#permissions
Reprints Information about ordering reprints can be found online:
http://www.neurology.org/misc/addir.xhtml#reprintsus

Neurology ® is the official journal of the American Academy of Neurology. Published continuously since
1951, it is now a weekly with 48 issues per year. Copyright . All rights reserved. Print ISSN: 0028-3878.
Online ISSN: 1526-632X.